Lund University Faculty of Science. STM-based characterization of single GaInP photovoltaic nanowires

Transcription

1 Lund University Faculty of Science STM-based characterization of single GaInP photovoltaic nanowires Author: Johannes Brask Supervisor: Rainer Timm Co-supervisor: Magnus Borgström Bachelor thesis Div. of Synchrotron Radiation Research May 2016

2

3 Abstract The I-V and photovoltaic properties of III-V semiconducting GaInP nanowires have been studied using a scanning tunneling microscope (STM) in evaluation for usage as a potential sub cell in a nanowire based tandem solar cell configuration. This evaluation required precise I-V characterization and photoresponse measurements of individual nanowires. Such measurements usually involve rigorous sample preparation, establishing metal contacts on both ends of the nanowire, which might have an inhibiting effect on the measured currents. In this work another approach has been employed allowing reproducible measurements on single upright standing nanowires. Using a standard STM, Ohmic contacts have been established with the Au seed particles of the GaInP nanowires. This reduced negative effects from the metalsemiconductor interfaces and required no sample preparation. In this way successful I-V characterization was conducted with currents in the µa range resulting in an average ideality factor of 2.04±0.08. Through laser illumination photocurrents of 19.3 pa were reached and a maximum open circuit voltage (V OC ) of 0.98 V. Furthermore this thesis presents measurements of the band gap (1.95 ev) and atomic imaging of the surface structure of GaInP nanowires which have not yet been reported in literature. These results could be useful in the creation of a nanowire based tandem solar cell with GaInP (with a large measured band gap of 1.95 ev) as a potential top cell in a tandem configuration with InP (1.34 ev) which has been studied to a great extent in the past. The attained results could also be of interest for further studies of GaInP nanowires in other purposes.

4 Acknowledgments I would like to express my sincere gratitude to my supervisor Dr. Rainer Timm for helping me establish the framework of this project and for being supportive throughout the entire process. With advice and reasoning he has been able to focus my efforts in the right direction, guiding me through the many theoretical and experimental conundrums of this work. Furthermore I would like to thank Jovana Colvin and Sofie Yngman for assisting me during the experimental process. They have shared many challenging hours in the lab with me and are both to thank for the success of this project. Special thanks to my co-supervisor Magnus Borgström for insightful conversations regarding the future of solar cell technology and to Olivier Scholder for saving me hours through correct data processing.

5 Contents 1 Introduction 1 2 Theoretical background Semiconducting materials p-n junction Diodes Ideal diode equation The ideality factor n Photocurrent I L Semiconductor applications; the solar cell III-V semiconductors NW array solar cells Scanning tunneling microscope (STM) STM-schematics Bias voltage Scanning tunneling spectroscopy (STS) Experimental process GaInP NW-array sample STM and STS on laterally deposited NWs NW deposition and cleaning STM and STS Top contact method for standing NWs Photocurrent measurements Measurements and results STM and STS results on a GaInP NW surface NW topography Atomic-scale surface structure Band gap and doping features I-V -characterization of upright standing GaInP NWs Ideality factor Photocurrent measurements I-V curves Open circuit voltage V OC and short circuit current I SC Discussion Interpretation of results Band gap and dopant properties I-V characterization of upright-standing NWs Ideality factor n Photocurrent Experimental process Conclusion and Outlook 28

6 Abbreviations CB Conduction band GaInP Gallium Indium Phosphide InP Indium Phosphide LDOS Local density of states NW Nanowire, sometimes refered to as wire PV Photovoltaic SEM Scanning electron microscope STM Scanning tunneling microscope UHV Ultra high vacuum VB Valence band

7 1 Introduction Ever since the first commercialized solar cells were introduced by Bell laboratories in 1954 with 6 % efficiency the market has expanded with an increasing rate due to the necessity for renewable energy sources. The most common planar solar cell material in use today is silicon, which at a relatively low cost of production dominates the industry. These industrialized solar panels convert about 15% of solar energy to electricity. In search of higher efficiency solar cells, the usage of III-V semiconductors (compounds of elements from the third and fifth group of the periodic table) have been exploited. By combining elements in this fashion semiconductors with direct band gaps can be created with higher absorption capabilities. Each semiconductor is however limited to efficiently absorb light with energy equal to (or slightly larger than) the band gap energy. By adding two or more semiconductors in one device a multi-junction or tandem solar cell can be made with the possibility of harvesting a larger portion of the solar spectrum. Such devices have under concentrated illumination reached conversion efficiencies of over 40% [1]. The cost of production for tandem solar cells are however limiting their possibility for industrial applications to fields such as space technology and terrestrial concentrator photovoltaics (CVP) [2]. Another limiting factor is the lattice-mismatch between different semiconducting materials and the strain caused to the final device due to the varying thermal expansion rates of the different sub cells. A possible solution to these problems has emerged through the expanding field of nanotechnology. III-V Semiconducting nanowires (NWs), from compounds such as GaAs, InP or GaInP can be grown on cheap Si substrates [3] creating solar cell devices of high flexibility using only a fraction of the material of similar planar devices. This relatively new field of research is showing great promise for next generation solar cell architecture. Initial attempts to create single junction NW array solar cells have resulted in low efficiencies of a few percent, in 2013 however Walletin and others reported InP NWs to achieve efficiencies up to 13.8% [4]. To reach efficiencies compatible to the highly expensive planar multi-junction solar cells researchers are investigating the possibility of creating tandem NW array solar cells of axially stacked III-V semiconductors connected via an Esaki-diode [5]. The success of such configurations is however dependent on precise characterization of the various semiconductors that make up these structures. Due to their large surface to bulk ratio NWs are expected to behave differently than larger bulks of the same material. To fully understand the behavior of a NW based solar cell it is therefore necessary to conduct measurements on single NWs. Due to the scale of these nano-structures this is by no means trivial and often requires significant experimental efforts. By establishing metal contacts to each end of an isolated NW [6] it is possible to create an Ohmic contact and preform I-V and photovoltaic measurements. This however requires rigorous sample preparations and is by no means efficient for contacting several NWs. Even with perfect surface conditions of the NW the metal-semiconductor interface between the NW and the contacts could create barriers for the generated current. The aim of this project is to study the I-V and photovoltaic properties of III-V Gallium Indium Phosphide (GaInP) NWs in a reproducible fashion without the use of metal contacts. This can be achieved with a standard scanning tunneling microscope (STM) using the top contact method [7]. The atomically sharp W tip of the STM can create metal-metal contacts with the Au seed particles of the upright standing NWs directly from the growth sample. In this way I-V characterizations of several NWs can be conducted without the barrier effects from metal-semiconductor interfaces. By shining a laser to the sample this process can also be used to measure the photoresponse of the NWs The GaInP NWs are expected to have a larger band gap than similar InP NWs (1.34 ev ) and could potentially be used as a top cell in a tandem configuration with InP as the bottom or middle cell. In order to investigate the size of the band gap of the NWs the STM can be used to perform tunneling spectroscopy (STS). This would also allow the possibility of investigating the location of the various doped segments of these NWs which is important for future creation of tandem configuration via tunneling diodes. 1

8 Through the attained I-V and photovoltaic properties of the GaInP NWs this project aims to conclude the possibility of using this material in a NW based tandem configuration. These wires were grown with ratios designed for InP NWs and the attained results in this work is not expected to break any records regarding the efficiency of GaInP NW-based solar cells. 2

9 2 Theoretical background This chapter will include a short description of the theory needed to understand the results and methodology of this work. 2.1 Semiconducting materials Although there is no clear definition of a semiconductor it is often refereed to as an insulators with a band gap less than 3 ev [8]. In comparison to a metal the Fermi energy E F is located within the band gap of the material. By introducing impurity or dopant atoms to a semiconductor their conductive properties can be controlled making them very useful in technological applications. The dopant atoms used in p-type and n-type doping is referred to as acceptors for p-type and donors for n-type p-n junction By combining p-type and n-type doped semiconductors a p-n junction or diode can be created. This junction will act as potential barrier for the generated charge carriers. When an n-type and p-type doped region are joint together the electrons in the CB on the n side will migrate to the p side while the holes in the VB of the p side will migrate to the n side. These charge carriers will recombine in the depletion layer, a region where there are no mobile charge carriers. This will cause the charge neutrality in the n and p doped areas within the depletion layer to vanish. The positively charged donor atoms and the negatively charged acceptor atoms will create an electric field acting as a potential barrier for the diffusion current I diffusion (see figure 1). The magnitude of this potential barrier is called the built in potential V bi. There is also always a small leakage current or drift current (I drift ) which passes the p-n junction, generated by minority charge carriers. By applying a bias over the junction the potential barrier in the depletion layer can be reduced (positive bias) or enhanced (negative bias). In this way the magnitude of the diffusion current through the circuit can be controlled. Figure 1: p-n junction with no external bias showing the diffusion current and drift current in opposite directions. In certain applications of the p-n junction such as in photo-voltaic devices it is favorable to insert an intrinsic (non-doped) region in between p an n creating a p-i-n or PIN diode. In doing so the capability of light absorption increases. If a photon of sufficient energy (larger or equal to the band gap of the semiconductor) is absorbed in the intrinsic region it will generate a electron hole pair which will join the majority charge carriers in the p and n parts (see figure 2). A solar cell is exposed to a great deal of light causing an accumulation of charge carriers in the p and n doped regions. Due to the potential energy of these charge carriers a current can be generated through the device. 3

10 CB VB E F Figure 2: p-i-n junction. An incident photon is absorbed in the intrinsic region generating an electron hole pair. This repeated process creates a build up of charge carriers in the n and p segments. 2.2 Diodes Ideal diode equation The behavior of a diode is in a simplified way explained by the ideal diode equation, also known as the Shockley ideal diode equation: I = I 0 (exp( qv ) 1) (1) k B T Where I is the net current through the diode, I 0 the dark saturated current (the leakage current when there is no generated photocurrent), V the applied bias and T is the temperature in Kelvin (see figure 3). This simplification does not take into account the Ohmic behavior of the current for large positive bias values nor the reverse break down phenomenon at large negative bias values The ideality factor n Another effect which is excluded from the ideal diode equation is the regeneration of charge carriers in the p-n junction. The amount of imperfections can be accounted for by introducing the ideality factor n into the equation: I = I 0 (exp qv nk B T 1) (2) In the ideal diode the magnitude of n is 1. In reality n always exceeds 1 and the magnitude of n determines the quality of the diode. To attain the ideality factor from measured I-V curves some simplifications are usually done regarding the equation. The 1 at the right most side is only significant at low values for the bias V (above 0.1 V the magnitude of the exponential increases rapidly) and the magnitude of I 0 is negligibly small. We thus reach the proportionality: ln(i) qv nk B T (3) This means that n can be attained by creating a semi-logarithmic plot of the measured I-V data. The ln(i) values will show a linear increase in the region where the I-V curves show an exponential increase. The slope m of the linear part of the semi-logarithmic plot can be used to determine n from equation 3: n = q mk B T (4) 4

11 2.2.3 Photocurrent I L Photocurrent or light generated current I L is generated by the separation of charge carriers through photon absorption as shown in figure 2. It can be thought of as a shift in the ideal diode curve. The direction of I L is opposite to the diffusion current and the curve is therefore shifted to negative current values (see figure 3). To calculate the total current I, the light generated current is subtracted from equation 1: I = I 0 (exp( qv k B T ) 1) I L (5) This shift of the current gives rise to two important parameters that characterize the performance of a solar cell. The current through the circuit without any applied bias is called the short circuit current I SC. At this point all the generated charge carriers will be used as current which will short circuit the device. The open circuit voltage V OC is the value of the bias voltage when the light generated current (in combination with the drift current) is equal in magnitude of the diffusion current (at I=0), determining the maximum value for the potential difference over the p-n junction generated by the charge carriers. Together the magnitude of I SC and V OC reveals the potential power output of the solar cell. The magnitude of the maximum power output P max can bee seen as a square in figure 3. Figure 3: The shift of the current through an ideal diode as a result of light exposure with the photo current in a dashed line. The maximum power output of a solar cell is marked P max and the dark saturated current I 0 is seen at large reverse bias values. 2.3 Semiconductor applications; the solar cell III-V semiconductors Pure intrinsic semiconducting elements such as Si or Ge have many properties which makes them suitable for a wide range of industrial applications, such as transistors, LEDs and solar cells. They however share the disadvantage of an indirect band gap. By combining elements from the third and fifth group of the periodic table, so called III-V semiconducting compound can be created. Some of which have direct band gaps which increases their ability to absorb or emit light, making them useful in opto-electrical devices. In terms of solar cells, these single junction devices are limited to absorb light of energy which correlates to the size of the band gap. With a large band gap (GaInP) high energy charge carriers can be generated, but this limits the range of absorption of the solar spectrum to light with short 5

12 wavelengths. With a small band gap (InP) light from a larger portion of the spectrum can be absorbed, however if the energy of the light exceeds the energy of the band gap the extra energy of the charge carriers will be lost to heat. Therefore the energy of the generated charge carriers is limited to energies slightly larger or equal to the band gap. On the basis of these restrictions Schockley and Queisser, in 1961, developed a theoretical limit of the efficiency of a solar cell, with a optimal band gap of 1.1 ev at 30 % [9]. Today with the possibility of adding III-V semiconductors in so called multi-junctions, this limit has been exceeded with measured efficiency of over 40 % [1]. These devices are however highly expensive and the cost of production makes them unable to compete with the abundant and relatively low cost silicon solar panels. Another problem is the varying rate of thermal expansion of the different materials in the junction causing strain in the device NW array solar cells Nanotechnology offers a solution to reduce cost and increase flexibility and absorption capabilities of highly efficient III-V multi-junction solar cells. Semiconducting NWs are one dimensional rod like crystal structures with varying diameters of nm with heights a few µm. They are applicable to fields such as electronics, photonics (LEDs) and are showing great promise in the field of photovoltaics. By growing homogeneous arrays of upright-standing sub-wavelength nanowires (NWs) of III-V semiconductor compounds, solar cells can be created using only a fraction of the material. In Figure 4 part of such an array is illustrated. The figure shows an effect known as resonant light trapping which increases the absorption capabilities of these devices. Classical ray optics can not explain this phenomena which is caused by the sub-wavelength diameters of the NWs. The incoming light resonates with the structure and if the diameter of the NWs is too small the light leaks out. By adjusting the diameter of the wire, optimal absorption coefficients can be found correlating to light of a certain wavelength. In this way NW array solar cells of various materials can be custom made to efficiently absorb light [4]. Light wave nanowire substrate Figure 4: Resonant light trapping in a NW array solar cell. The current challenge in the development of NW array solar cells is the desire to construct multijunction NWs of several III-V semiconducting materials. It is therefore important to study III-V compound NWs with varying band gap sizes. The efficiency of these devices is thought to compete with the highly expensive planar III-V multi-junction solar cells at fraction of the material cost. 2.4 Scanning tunneling microscope (STM) The scanning tunneling microscope (STM) was invented by Gerd Binning and Heinrich Rohrer in 1982, an achievement for which they received the Nobel prize four years later [10]. This new invention allowed scientists to successfully visualize electronic properties of single atoms and atomic structures. It is based on the principle of quantum tunneling and uses piezo electronics to control the fine (sub nm scale) motion of an atomically sharp tip which enables topographic imaging of atomic structures and local I-V measurements. 6

13 Tunneling current To understand the concept of tunneling current a quantum mechanical approach is necessary. The electrons in the atoms of a solid (in this case the sample) have lower energy than free electrons in the vacuum and thus remain in the atom. However if another solid (the tip) is present at a finite distance from the sample the electrons in the atoms of the sample have a finite probability to tunnel through the vacuum barrier to the tip. The probability for an electron to be at distance z from equilibrium position is described by the square of the wave function ψ(z) 2 and depends on an exponential decay of the wave function in the following way: ψ(z) 2 = ψ(0) 2 e 2kz (6) where k is the decay rate, defined as: k = 2m e (φ ev T ) h 2 (7) φ is the work function which describes the minimum energy needed to remove an electron from a certain atom, ev T is the applied sample bias and m e is the electron mass. In the case with the sample and the tip the probability of an electron to tunnel through the vacuum barrier would be the probability of finding the electron at the edge of this barrier, i.e at the tips distance d to the sample atom. From this the transmission coefficient can be defined [11]: T = ψ(d) ψ(0) e 2dk (8) The transmission coefficient effectively describes the proportionality of the tunneling current to the distance d between the tip and the sample: I e 2dk (9) This shows that the current is strongly dependent of the distance between the tip and the sample. Small changes in d creates an exponential increase or decrease in the tunneling current. These swift changes in the current signal allows for the high sensitivity needed to perform movements in the pm range with the STM. It also restricts the current to atoms on the surface layer of the sample which is required for imaging with atomic resolution. The fact that the current is dependent of the distance to the sample is however a simplification. It is the local density of states (LDOS) of the surface atoms that govern the magnitude of the tunneling current. To attain a direct proportionality for the tunneling current the LDOS is integrated from the Fermi level E F to the applied bias at E F +ev t [12]. This is explained by the following relation: evt I ρ t ρ s,loc ( r 0, E F + ɛ)dɛ (10) 0 where ρ t is the LDOS for the tip and ρ s,loc ( r 0, E F ) the LDOS for the sample surface at the tip position r 0 and energy E F STM-schematics Figure 5 below displays a schematic overview of the setup of the STM. The tip, usually made of tungsten, is ideally one atom wide at the top. In the figure there is a tunneling current generated through the top most atom of the tip and one single atom of the sample. This case is ideal, usually the tip gets blunt through collisions with the sample and the current is not only generated between two atoms. The analog signal of the tunneling current is converted into a voltage signal in the tunneling current amplifier. This signal is compared to a chosen reference value and the difference controls the motion of the tip in the z-direction (up and down) through a constant feedback loop. The distance control unit applies a voltage to the z-piezo which keeps the tip at a equilibrium z position. The x an y piezos control the motion of the 7

14 tip in the x,y-plane where the chosen area is scanned line by line in a sweeping manner. As the tip follows the contours of the sample surface (the LDOS of the sample atoms) it is pushed towards the sample (at surface depressions), which is represented as darker areas in the topographic image, or away from the sample (protrusions) which is represented as brighter areas. This process is called constant current mode. Figure 5: Schematic overview of the basic constituencies of the STM. Figure: Michael Schmid, TU Wien 2.5 Bias voltage In most cases the bias voltage is applied to the sample, leaving the tip grounded [13]. The sample bias voltage can be used to control the direction of the current. If V > 0 the electrons will tunnel from the tip to the sample and vice versa. The resulting image will be either that of the empty or occupied states of the atoms in the sample. Figure 6 below shows the effect of the polarity of the sample bias when scanning a semiconducting material. With a positive sample bias the Fermi level of the tip is above the CB of the sample which allows electrons to tunnel from occupied states in the tip to empty states of the sample (empty-state STM). Filled-state STM is instead when the sample bias is negative and E F of the tip is below the band gap of the sample. With no applied bias E F of the tip is in the middle of the band gap between the valence band and the conduction band. In this case no current can be generated. If the applied bias voltage is too small there will be no tunneling current which will cause the tip to collide with the sample. CB Tip Vacuum CB E F VB Sample Tip I Vacuum CB E F VB Sample Tip I Vacuum Sample E F VB Figure 6: The effects of the direction of the sample bias (of magnitude ev ) on the tunneling current. a) No sample bias. b) Positive sample bias, tunneling current from the tip to the sample. c) Negative sample bias, tunneling current from the sample to the tip. 8

15 2.5.1 Scanning tunneling spectroscopy (STS) Another useful tool of the STM is the possibility to perform tunnel spectroscopy. In this case the z position of the tip is frozen by turning off the feedback loop. A sweep of the bias voltage is then executed within a chosen range. By running the sweep between positive and negative values the size of the band gap of a semiconducting material can be measured. Figure 7a shows a sweep between negative and positive sample bias on the III-V semiconducting material GaInP. The segment around V =0 where the current is zero is where the Fermi energy of the tip is within the band gap of the sample. In order to study the LDOS at a specific energy ev T a sinusoidal voltage is added to the bias and any swift changes in the measured tunneling current is detected by a lock-in amplifier and a di/dv signal is directly attained during the measurements [12]. The result of which can be seen in figure 7b. (a) (b) Current(pA) di/dv signal(pa/v) VB CB Sample bias(v) Sample bias(v) Figure 7: STS on GaInP with a fixed tip position. a) The I-V curve. b) The derivative of the current with respect to the voltage plotted against the voltage (di/dv -V ). The CB and VB and indicated. The di/dv signal is however dependent on the distance from the tip to the sample (which might vary during spectroscopy measurements). This can be compensated for by normalizing the di/dv signal with the total broadened conductance I/V [12]. 9

16 3 Experimental process This chapter will give a presentation of the experimental processes used in this work and a description of the GaInP sample. In order to conduct spectroscopy over various parts of the NWs and atomic resolution imaging of the surface, the NWs were deposited on an InP substrate and cleaned. The I-V characterizations of the uprights standing NWs required the usage of the top contact method with the STM. The same method was used with the addition of laser illumination to measure the photoresponse of single NWs. 3.1 GaInP NW-array sample The sample used in this study consists of single junction III-V compound p-i-n doped Ga 0.46 In 0.54 P 1 NWs on a p-type doped InP substrate. Figure 8 shows an image illustrating the various sections of the NWs alongside a scanning electron microscopy (SEM) image of a part of the array. The top part is the Au seed particle. The diameter of the NWs is 188 nm and the height is 1.7 µm. The p and n-type doped parts are 320 nm long respectively, the intrinsic region is deliberately much longer at 900 nm. The length of the intrinsic region is (as discussed above) of utmost importance for a PV-device since the photo current is generated there. The array is grown on an InP substrate using metal-organic vapor phase epitaxy (MOVPE) in a hexagonal pattern. The distance between each NW is 500 nm, (from the center of each wire to the next). n i p Figure 8: a) Illustration of the GaInP NWs studied in this thesis work displaying each region of the NW. b) SEM image of the GaInP array. Courtesy of Vilgaile Dagyte. SEM-imaging of the sample showed an unwanted defect from the growth-process. Some of the NWs are much longer than the desired height of 1.7µm. The distance between these defect NWs are however sufficiently long not to cause any major difficulties in the process of scanning the sample with the STM. 3.2 STM and STS on laterally deposited NWs To be able to study the various sections of p-i-n doped NWs, using STM/S, they must be deposited on a new surface where they will be lying flat. This will enable surface characterization over the entire length of the wires. 1 Showing the Gallium-Indium ratio in the compound with 46% Ga and 54% In 10

17 3.2.1 NW deposition and cleaning The deposition is a mechanical process and the results may vary a great deal. An efficient way to exchange substrate host for the NWs is to simply place the growth sample and a suitable substrate in contact with each other [12],[14]. By gently applying a force the standing NWs break off from the substrate of the original sample and stick to the new surface. It is of great importance that the NWs break at the substrate and not further up on the wire. This might cause one of the doped regions to be excluded, which is especially likely for NWs constructed for photovoltaics since the p and n-type doped regions are very short. Figure 9 displays GaInP NWs deposited on a InP substrate (NWs are bright and the substrate is dark). 9b shows an area where too many NWs have been deposited in great bundles. This is likely the effect of applying too much force when contacting the surfaces. The desired spread of NWs should resemble that of figure 9c where NWs could be singled out and easily located. Figure 9: Laterally deposited GaInP Nws on an InP substrate, courtesy Sophie Yngman. a) Edge of the InP piece where the region of deposited NWs are located. The NWs are bright as apposed to the dark substrate. b) Area with a bundle of NWs stacked together. c) The edge of the deposited NW region containing a good spread of individual NWs (bright) and some minor bundles. Successful surface cleaning is crucial for measuring atomic scale surface properties. To remove contamination the sample is heated in the STM in UHV. This process is however limited for III-V semiconductors since they will be damaged at too high temperatures. By combining the heating method with the introduction of Hydrogen atoms, which binds to the oxides in a advantageous manner, it is possible to effectively clean the sample at lower temperatures [12]. This can be done with a hydrogen source and a hot filament, which thermally cracks the hydrogen gas to (H ) STM and STS In order to conduct STM and STS measurements on single deposited NWs they must be located using the STM tip. This process is rather difficult considering the diameter of these wires (over 100 nm). Since the STM is designed to study surface structures with a z-range of a few Å, a couple of measures are needed to assure that the tip does not collide with the NWs. The tunneling current is decreased (which increases the distance to the sample) and the speed of the feedback is increased. By roaming the tip over the substrate and simultaneously studying the movement of the tip in the z direction the NWs can be located. To conduct spectroscopy a flat surface on top of the wires is chosen. By increasing the tunneling current and decreasing the speed of the feedback imaging of atomic resolution of the NW surface is also possible Top contact method for standing NWs There are several methods of creating Ohmic contacts to single NWs, however the usage of a standard STM has proven very effective [7]. The top contact method of the STM allows measurements of several NWs in the array without sample processing. The metal-metal contact of the STM tip and the seed particles of the NWs decreases the metal-semiconductor interface which is problematic in other methods 11

18 such as the usage of metal contacts. The ultra high vacuum (UHV) chambers of the STM also allows control over the surface conditions of the sample and the probing tip. After the tip has been automatically approached to the sample it is retracted a number of steps to assure that it is above the top of the NWs. The tip is then extended fully and a given area (large enough to contain a few NWs in the array) is scanned in the constant current mode. Since the tip is fully extended, even though the constant current parameter is not reached, the tip can not approach the sample further. At this point there is no recorded tunneling current since the NWs are too far away. Approaching step-wise (in varying steps of a few hundred nm) and scanning a portion of the area large enough to conclude that the tip is still too far away, the process continues until the tip probes a tunnel current at the top end of the NWs. The tip should be no more than 100 nm below the top of the NWs (ideally less) to ensure that there are no collisions. This is rather difficult to achieve and the process takes some time to master. After which a topographic map of a few NWs in the array (see figure 10) can be taken. Figure 10: STM image of an array of standing GaInP NWs. After this, the tip is positioned above the center of one of the NWs. The feedback loop is turned off leaving the position of the tip fixed above the chosen NW. The tip is then forced a few nms into the metal particle on the top of the NW in a controlled fashion establishing a point contact. An external voltage source is then applied between the tip and the NW substrate to conduct a bias sweep. Using a current preamplifier the I-V properties of the NW can be measured. To improve the point-like Ohmic contact the tip is pushed further into the NW in small steps of 5 nms and a I-V measurement is conducted for each new position. The process is repeated until the optimal contact is reached, as described in [7]. 12

19 Nanowire Tip Nanowire Tip V a) b) Feedback loop ON Feedback loop OFF Substrate Substrate Figure 11: Top contact method of the STM. a) Tip scanning the NWs, feedback loop is on. b) Tip has been pushed into the top of a NW, the feedback loop is off and an external voltage has been applied Photocurrent measurements Through the windows of the STM it is possible to expose the sample to various light sources at different angles. The top contact method is used to locate the NWs and to establish an optimal contact (as described above). In this case however it is important not to push the tip too far into the top of wires which might damage them. It is therefore necessary to use shorter steps of 1 or 2 nm when the magnitude of the measured current increases less drastically. If the optimal contact is acquired the resulting current will not increase at all if the tip is moved further in. The process as a whole requires some experimental trial and error. Once the optimal contact is attained a bias sweep is conducted. The light source is turned on during the sweep, exposing the sample. Since the I-V curves might differ if several sweeps are conducted on the same NW it is relevant to measure the dark and photo generated I-V curves during the same sweep. In this way a reference current is produced for each measurement of the photocurrent. To control the quality of the setup, the direction of the light can be adjusted during a bias sweep. It is possible to improve the direction by studying the effects on the photocurrent which is displayed on a computer during the sweep. 13

20 4 Measurements and results This chapter will present the conducted measurements and attained results from the experimental work of this thesis. This includes STM/S measurements of the surface on deposited GaInP NWs, I-V characterization of upright standing GaInP NWs from the sample array and photocurrent measurements with an estimation of the open circuit voltage V OC. 4.1 STM and STS results on a GaInP NW surface The GaInP NWs were deposited on a InP substrate (see image 9). The substrate with the NWs was scanned by STM with a bias voltage of -2 V. The process of localizing the NWs (described in section 3.2.2) proved difficult as they appeared to attach to the tip during scanning. This might have been caused by oxides between the NWs and the surface which had not been removed during the cleaning process. The sample was prepared and cleaned three times using the Hydrogen source. Each time new parameters for the temperature and exposure time was used until the NWs no longer detached from the substrate during scanning, allowing for STM imaging of the surface and STS measurements. The final cleaning parameters are presented in table 1. Table 1: The cleaning parameter for GaInP NWs deposited on InP(111)B substrate. The hydrogen source was heated to 1700 C and the hydrogen gas was thermally cracked to (H ). The pressure was kept at 10 6 mbar. GaInP NWs Attempt 1 Attempt 2 Attempt 3 Temperature C Cleaning time (min) All STS measurements were conducted on the NW shown in figure 12. To estimate the size of the band gap linear fits were made on the (di/dv )/(I/V ) curves as explained in section For a large portion of the attained data such a fit was not possible due to the unexpectedly broadened shape of the curves and the non linear behavior of the current outside the band gap NW topography Figure 12 displays the NW on which all STS measurements were conducted. The width of the NW in the image greatly exceeds that of the actual wire (which should not have been affected during deposition). This can be attributed to the bluntness of the tip used for scanning. What is of most interest is the length of the wire. In the image it is 2µm which is a good indication that the majority of the NW, with the known length of 1.7µm (see image 8), has been transferred during the deposition process. The additional 0.3 µm in the image could be a result of the condition of the tip. Any precise approximations on the actual length of the deposited wire in the image is not possible since the shape of the tip is unknown. The bright vertical lines in the center of the image is not a feature of the NW but noise caused by some atom(s) which the tip might have picked up during scanning. This can create a swift increase in the tunneling current causing the tip to retract to avoid collision. The tip would then slowly return to the surface creating bright lines in the image in the direction of where the tip is scanning. 14

21 Figure 12: STM image of a GaInP NW deposited on an InP substrate Atomic-scale surface structure Figure 13 displays a topographic image taken on the middle region of the GaInP NW displayed in fig 12. Since a negative sample bias was used (filled state STM), the bright dots in the image represent areas with a high LDOS, which in the case of filled state imaging are located at the P atoms of the sample. The surface structure is most likely zink blende{110} which is characterized by the distinguishable diagonal rows of the surface atoms [14]. Figure 13: STM image of the surface structure of the GaInP NW displayed in figure 12. The atomic resolution is a good indication that the cleaning process of the NWs was successful. Even though contamination elements were discovered during other surface scans there are areas like the one displayed in figure 13 that have been cleaned from all contamination. The fact that the tip used in the scanning process was blunt (same tip as in figure 12), is remarkable considering the atomic resolution of the attained image Band gap and doping features After the surface condition of the NW in figure 12 was investigated, tunneling spectroscopy measurements were conducted in various locations of that wire. In total ten measurements of the size of the band gap were conducted with a calculated average value of 1.95±0.15 ev. Figure 14 contains the (di/dv )/(I/V ) curves of one measurement. The up and down sweeps, from negative to positive bias values (a) and positive to negative (b) are plotted separately and the average curve is plotted in black. 15

22 (a) (b) (di/dv) / (I/V) (a.u) Sweep 1 Sweep 2 Sweep 3 Sweep 4 Sweep 5 Average sweep (di/dv) / (I/V) (a.u) Sweep 1 Sweep 2 Sweep 3 Sweep 4 Average sweep Sample bias (V) Sample bias (V) Figure 14: STS conducted a GaInP NW. The two plots contain the (di/dv ) / (I/V ) curves from the up sweeps in (a) and down sweeps in (b). The average curves are plotted in black. Linear fits were created to the average curves from figure 14. The magnitude of the band gap for the up and down sweeps were 2.00 and 1.71 ev respectively. These linear fits are presented in figure 15. The difference in the size of the band gaps for the up and down sweeps of the same measurements could be explained by contamination on the NW surface. This is discussed further in section (a) (b) 5 3 (di/dv) / (I/V) (a.u) Average sweep Linear fit (di/dv) / (I/V) (a.u) Average sweep Linear fit Sample bias (V) Sample bias (V) Figure 15: Estimations of the band gap for the average curves of the sweeps presented in figure 14 with the linear fits in red. Estimation of the band gap for the up sweeps (2.00 ev) are shown in (a) and the down sweeps (1.71 ev) in (b). As the bias sweeps over voltages from negative to positive (up) and from positive to negative (down), there is a slight shift in the attained current values. This is an experimental defect whose magnitude is dependent on the chosen time constant of the lock in amplifier in the set up. The lock in amplifier sums up an average of measured current values over a certain time period causing a shift of the entire curve. Figure 14a shows five sweeps from negative to positive sample bias values (up) and an average sweep (in black). The curves are clearly not symmetric around V=0 but shifted to positive bias values. Figure 14b shows the down-sweeps of the same measurement, here the curve is shifted to negative bias values. Since the two curves are shifted by the same amount the band gap is concluded to be roughly symmetric around zero placing the Fermi energy at equal distance from the CB and VB. This means that 16

23 the measurement was conducted on the intrinsic region (no dopant atoms) of the NW which correlates to the fact that it was taken on the middle region of the NW in figure 12. Several attempts were made to identify the p and n doped regions of the NW in figure 12. Although there were isolated examples of measurements were the up and down sweeps were shifted to either side of V=0, they were not reproducible. Figure 16 displays one such case. Both the up and down sweeps produced curves which are both shifted toward negative current values. This would indicate an n-type doped region where the Fermi energy is closer to the CB. Although more measurements were conducted in the same region the resulting curves did not concisely indicate any shift of the band gap energy. (di/dv) / (I/V) (a.u) a) Sample bias (V) (di/dv) / (I/V) (a.u) b) Sample bias (V) Figure 16: STS conducted on the left part of the NW in figure 12. (a) contain the up-sweeps and (b) the down sweeps. It is apparent that both curves are shifted to the left indicating an n-type doped region with the Fermi energy closer to the CB. The main results in this section is the attained value for the band gap of 1.95±0.15 ev. This value is greater than for similar InP NWs (1.34 ev) [4]. These GaInP NWs could thus be relevant as a top cell in a tandem configuration with InP. 4.2 I-V -characterization of upright standing GaInP NWs This section presents the attained result from measurements conducted on the upright standing GaInP NWs using the top contact method of the STM. I-V characterization of 15 NWs were conducted using two separate W tips. The NWs were located using the constant current mode and the top contact was established (as explained in chapter 3.2.3). For each wire two bias sweeps were conducted from -1 to +2 Volts. To attain the ideality factor n semilogarithmic plots were made from the measured currents. The values for the slope m of the linear part of the semi-logarithmic plots were determined using a linear fit (first degree polynomial) for a V range within the linear region for all measured wires (the range was established simply by studying the curves). The ideality factor was calculated using equation 4, with the temperature of the sample assumed to be at 293 K (room temperature). Due to failure of the hydrogen source the sample was not cleaned in any way in preparation for these measurements and for the photo current measurements presented in chapter 4.3. I-V -curves Figure 17 displays the attain I-V curve for NW 8. The exponential increase of the curve is visible as the linear region between 1 and 1.5 V in the semi-logarithmic inset. This concludes that this NW show a good diode behavior. The low currents at negative V shows that there is a very low leakage current. This is the case for all measured wires and can be clearly seen in figure 19. The maximum current (at 17

24 2 V) for wire 8 is 2.35 µa is large considering the limitations the small diameter of these NWs have on the current density. The inset also indicate that Ohmic resistance is reached at roughly 1.5 V, where the diode of the NW no longer dictate the magnitude of the current through the circuit Wire Semi-log. plot Current (A) Sample bias (V) Figure 17: I-V curve for nanowire 8 of the GaInP sample. The figure contain a semi-logarithmic inset of the I-V curve between 1 and 2 V. All measured I-V curves are presented in two plots in figure 18 since two different STM tips were used. Wires 1-8 were measured with a blunt tip and wires 9-15 were measured with a newly prepared sharp tip. The highest measured current values (at 2 V) is continuously larger for wires 1-8 with one exception (wire 7). These values varied from 0.66 µa (wire 9) to 5.22 µa (wire 2). The large difference in the spread of the maximum current values is clearly seen in figure 19. Here all measured wires are plotted with NWs 1-8 in red and NWs 9-15 in black. The figure also contains an inset with the current-values for negative and low positive bias values. These values are constant within a range of 0.1 na. The current is not, as in theory, purely negative at these voltages which is an experimental defect. At zero voltage the Ampere-meter used to measure the current is set to zero but a small fluctuation in the measured current still exists. The initial increase (or dominance) of the diffusion current is seen at 1 V. 18

25 (a) (b) Current (A) Wire 1 Wire 2 Wire 3 Wire 4 Wire 5 Wire 6 Wire 7 Wire 8 Current (A) Wire 9 Wire 10 Wire 11 Wire 12 Wire 13 Wire 14 Wire Sample bias (V) Sample bias (V) Figure 18: I-V curves for nanowires 1-15 of the GaInP sample. a) Wires 1-8, measured with a blunt tip (wire 8 is measured with a shorter step size for the sample bias). b) Wires 9-15, measured with a sharp tip Wire 1-8 Wire Current (A) Sample bias (V) Figure 19: I V curves for NW 1-8 in red (blunt tip) and 9-15 in black (sharp tip) of the GaInP sample. The figure also contain an inset with the current-values for negative and low positive bias values indicating a low leakage current. The different behavior of the I-V curves from wires 1-8 and 9-15 is further presented in figure 20 where the natural logarithm of the current for all wires is plotted against the voltage. Here the linear part of the curve is the exponential part of the I-V curve i.e where the magnitude of the current through the circuit is dictated by the p-i-n junction of the NWs. The semi-logarithmic plot shows this region clearly in contrast to the I-V curves in figures 18 and 19. Two important points can be drawn from the general behavior of the wires measured by the two tips: 19

26 There is a offset in the curves in the linear region. Wires 1-8 (red) are as mentioned above generally higher in current for a given voltage. Wires 9-15 have a slightly steeper increase in the linear region and thus a faster exponential growth of the current. This region is also shorter for wires This indicates that the contacts created with the sharp tip were of higher quality. The lower currents for these wires (9-15) is however an indication that the resistance for the sharp tip was greater than for the blunt tip. This will be discussed further in section Wire 1-8 Wire 9-15 Current (A) Sample bias (V) Figure 20: Semilogarithmic plot of wires 1-8 (red), measured with a blunt tip and wires 9-15 (black), measured with a sharp tip Ideality factor The measured values for the ideality factor n is presented in table 2 with an average value of The lowest n is found with NW 10 at The spread of the values for n between wires 1-8 (2.44±0.18) is significantly larger than the spread for wires 9-15 (2.04±0.08). The magnitude of n is also greater for NWs 1-8 with one exception in wire 8 (with n=2.09 ). Figure 21 display the semi-logarithmic plot of NW 8 with a linear fit between the sample bias values 1.15 V-1.40V. 20